identifying coastal forest merlin (falco columbarius suckleyi) breeding habitat using geographical...
TRANSCRIPT
Identifying Coastal Forest Merlin (Falco columbarius suckleyi) Breeding Habitat Using
Geographical Information Systems
Christopher M. Talley Western Washington University
Abstract. The Coastal Forest Merlin (Falco columbarius suckleyi) natural breeding habitat
has traditionally been located in the temperate rainforests of the Pacific Northwest of the United
States and Canada. The natural habitat has experienced a significant reduction in habitat quantity
due to anthropogenic influence over the last 150 years. Despite the reduction in their traditional
habitat the Merlin has been able adjust to changes in habitat and expand their population. During
each breeding season from 1986 to 2013, nest sites and home ranges were geographically located
for a population of Coastal Forest Merlin (Falco columbarius suckleyi) in an 116,660 km² study
area that encompassed areas of Northwestern Washington State and Southeastern and the Central
Interior of British Columbia, Canada. Geographical Information Systems (GIS) and satellite
imagery were used to determine and compare the amount, distribution, and configuration of
several key habitat variables within 8km circular plots centered on known nest sites and random
control sites. The plot size utilized for analysis was based on the observed distribution and
behavioral characteristics of Merlins within the study area Analysis demonstrated that complex
habitat edge configuration, greater spatial heterogeneity, and higher amounts of habitat richness
were the most significant Merlin breeding habitat variables. The results of analysis are intended
for landscape scale systematic analysis of breeding habitat variables, demographic assessments,
and to guide recovery and management decisions.
Key Words: Coastal Forest Merlin; Falco columbarius suckleyi; Raptors; Geographical
Information Systems, habitat use; landscape ecology.
INTRODUCTION
In 2012, the Coastal Forest Merlin was listed in Washington State as a species of
concern based on the declining amounts of suitable habitat throughout their range, declining
population trends, and lack of existing regulatory methods to protect the species (WDRW, 2012).
Habitat characteristics such as land cover, land use, vegetative composition, and spatial
configuration are key elements that influence wildlife breeding success and species evolution
(Rodiek & Bolan, 1991). In order to comprehensively understand the importance of land cover
characteristics on a species viability and development, it is necessary to effectively model
wildlife habitat in terms of spatial and land cover characteristics (Turner & Gardner, 1991). With
increased importance being placed on spatial dynamics in relation to landscape ecology and
species evolution, it important to utilize modeling methods that consider variety of
environmental variables to produce relevant habitat models (Tutle. Et. al, 2006). Quantifying
information about habitat characteristics such density and distribution over large geographical
areas by field surveys can be expensive, time consuming, and impractical. To complete this task,
computer derived habitat models created using Geographical Information Systems (GIS) can be
used effectively to depict the vegetative and groundcover characteristics of a large study area
1
(Lillesand and Kiefer, 1987). Geographic Information Systems significantly increase
productivity by allowing researchers to efficiently survey and analyze many aspects of wildlife
ecology without the spatial and temporal limitations of traditional methods (Shaw & Atkinson,
1990).
The purpose of this project was to create a GIS based landscape level habitat model to
evaluate breeding habitat characteristics of the Coastal Forest Merlin in a study area which
comprises areas of Northwestern Washington State and Southwestern British Columbia. The
model building process consisted of creating a digital database by compiling and interfacing
comprehensive digital vegetation and land cover data generated from Landsat 7 ETM+
multispectral data satellite imagery, vector and raster based political and environmental data
created by various agencies, and field gathered groundcover data.
The objectives of this study were to: 1) determine the quantity and spatial distribution of
different of habitat variables, 2) evaluate the relative importance of habitat variables on Merlin
distribution and abundance, and 3) investigate species demographics and density in different
habitat types. The information generated by this project is designed to provide knowledge and
guidance to help inform scientists, policy makers, and property owners in making prudent
resource management decisions that will help retain valuable habitat for the continued
reproductive success for the Coastal Forest Merlin.
Study Area
The habitat modeling described in
this report was conducted on an 116,660
km² (11,666,048 ha) study area spread
over portions of Northwestern Washington
State and Southeastern British Columbia
Canada. The study area was bounded
approximately by the geographical
coordinates of 128º W to 120 º W
longitude, and 46º N to 55º N latitude. The
land forms consist of highly developed
floodplains and coastal lowlands, heavily
forested and rugged coastal mountainous
regions, and drier inland moderate
elevation plateaus.
Land ownership is a broad mix of
public, private, and tribal owned lands.
The study area occupies parts of 4 distinct
ecologically and geographically defined
level III ecoregions; 1) the Puget Trough-
Georgia Basin, 2) the Pacific Northwest
coast., 3) the central interior of British
Columbia, and 4) North Cascade and
2
Pacific range (Nature Conservancy,2006).
The Puget Trough- Georgia
Basin ecoregion occupies a long narrow
continental glacial trough that consists
of many islands, peninsulas, and inlets.
Elevations in the area range from sea
level to 750-1000 meters in the foothills.
The area is characterized by a mild
maritime climate with mild, wet winters.
Summers are fairly warm and dry and
often overcast. Mean January
temperature is 4° C and mean July
temperature is 18° C. Precipitation,
falling primarily as rain, averages 100
cm per year. The Olympic Mountains
created rain shadow areas that include
the northeast corner of the Olympic
Peninsula, Whidbey Island, and the San
Juan Islands. The annual precipitation
tends to be lower in these areas
averaging 40 to 75 cm.. Rainfall is
higher in the foothills due to the
orographic lift created by the Cascade
Mountains and averages 150-200 cm a
year.
The natural landscape consisted of thick coniferous forests that grew on areas consisting
of glacial moraines, floodplains, and river terraces. Douglas-fir (Pseudotsuga menziesii), western
hemlock (Tsuga heterophylla), western red cedar (Thuja plicata), and grand fir (Abies grandis) are the predominate species in the upland forests, while black cottonwood (Populus
trichocarpa), red alder (Alnus rubra), and big leaf maple (Acer macrophyllum) are the common
forest elements in riparian areas.
The Pacific Northwest Coast region includes the coastal Ranges of Northwestern
Washington State and Vancouver Island. The region has landforms that consist of beaches, low
marine terraces, sand dunes, and spits in the marine areas, headlands, high marine terraces, and
low mountains in the uplands, and the lower portions of the Olympic Mountains up to around
1200m in elevation. The Coast Range’s climate is influenced by cool, moist air from the ocean.
Mean January temperature is 6° C and mean July temperature is 12° C. Precipitation falls mainly
as rain at the lower elevations and averages 150- 250 cm a year, with some areas receiving
upward of 500 cm of rain a year. The coastal lowlands and low mountains are dominated Mature
forest consist primarily of Coast Douglas Fir (Pseudotsuga menziesii var. menziesii), western red
cedar, western hemlock, and Douglas fir. Pacific silver fir (Abies amabilis) and mountain
3
Hemlock (Tsuga mertensiana) are the common forest elements at higher elevations. Wetter
and riparian areas supports red alder, black cottonwood, western red cedar, and big leaf maple.
The understory typically contains salmonberry (Rubus spectabilis), salal (Gaultheria
shallon), western sword fern (Polystichum munitum), vine maple, and Oregon grape (Mahonia
aquifolium).
The Cascade Mountains and Pacific range region is primary high mountainous area in the
study area consists. The alpine areas consist of glaciated mountain terrain with elevations up to
2000m, and several large composite volcanoes that rise to over 3000m. The region receives high
amounts of precipitation from 150 to 400 cm a year as rain or snow. The higher elevations can be
covered by as much as 6 m of snow in the winter.
The vegetation in the region is highly diverse. The mountainous areas have a moist,
temperate climate that supports an extensive and highly productive coniferous forest that is
intensively managed for commercial logging. At lower elevations, Douglas-fir, western hemlock,
western red cedar, big leaf maple, and red alder are typical. At mid elevations, Pacific silver fir,
mountain hemlock, noble fir (Abies procera), and lodgepole pine (Pinus contorta) are the
common tree species. A mosaic of mountain hemlock, Pacific silver fir, yellow cedar, and
subalpine parklands occurs at higher elevations. Disturbed areas can be lined with Sitka alder or
vine maple.
The interior of British Columbia ecoregion occupies a plateau in the central portion of the
province with long forested sections into the valley bottoms of mountainous areas to the north,
east, and west. Elevations range from 750 to 1500 m. Several major lakes and rivers are located
in this zone. The area experiences extremes of temperature; the summers are short with warm
temperatures that can reach a high of 30 degrees Celsius. Winters can reach temperatures of -10
degrees C, with extremes sometimes at -40 degrees C.
The rolling landscape of the Sub-Boreal zone is covered in primarily coniferous forest.
Pioneer species include the trembling aspen (Populus tremuloides) and paper birch (Betula
papyrifera) in the uplands. The dominant coniferous species are hybrid white spruce (Picea
glauca), subalpine fir (Abies lasiocarpa), and occasionally, black spruce (Picea mariana), along
with lodgepole pine and occasionally Douglas-fir. Primary components of the understory
include; Queen’s Cup (Clintonia uniflora), Devil’s club (Oplopanax horridus), Sitka alder
(Alnus viridis), and multiple species of wild berries.
Focal Species
The Coastal Forest Merlin (Falco columbarius suckleyi) is one of three North
American sub-species of Merlin Falcon. The sub-species inhabits the Pacific-Northwest
temperate coastal rainforests. They tend to nest adjacent to rivers and water bodies near forest
openings or edges (Johnsgard, 1990). The relatively small yet sturdy bird uses its speed and
agility to prey on small song and shorebirds, insects, and small mammals (Cade, 1982). Merlins
use the same general area year after year for breeding, but not necessarily the same actual site,
particularly if young were fledged the previous year (Brown & Amadon, 1968). Nest are almost
exclusively is located in high in mature conifer trees with a complex canopy structure (Sohdi et
4
al., 1993). The Merlin can occupy elevations that range from sea level to near the tree line.
Historically this subspecies has preferred breeding and foraging sites that have complex forest
structure common to late seral stage conifer forest. The subspecies native natural habitat of
temperate rainforest has experienced significant decline in quantity and continuity due to
anthropogenic influence over the last 150 years (Stillman, R.C., personal communication, August
26, 2014). Their conservation status varies by jurisdiction In Washington State they are classified
as a Species of Concern and as a State Candidate Species (WDFW, 2013). In Canada, the Merlin
is considered not at risk (COSEWIC, 2009). In light of the decline in natural habitat, the Merlin
is highly adaptable in terms of habitat selection and has proven successful while occupying nests
in more densely populated areas. Although the breed prefers mature conifer trees for its nest
sites, they have the ability to thrive in areas of intense human activity and forage in areas with a
high level of habitat class variability.
METHODS
Field Methods
The study area was subject to a long term demographic study conducted by the Merlin
Falcon Foundation from 1986-2013 during which Coastal Forest Merlin were non-invasively
monitored to establish distribution and relative abundance of the population (Drummond &
Stillman, 2014). The systematic searches occurred while assessing behavior and reproductive
status during each breeding season (roughly mid/late Feb to early August). During field surveys,
each nest site observation of a territorial resident of the population was documented and assigned
a unique identification code. Relevant nest site microhabitat, geographical, and environmental
characteristics were integrated with the breeding data and nest location information (Drummond
& Stillman, 2014)
In order to develop a multivariate set of habitat characteristics designed to analyze avian
associations and habitat use, a GIS dataset defining Merlin activity centers, as well as relevant
geographic and environmental data, were created using ESRI Arcmap 10.2 (ESRI 2014. ArcGIS
Desktop: Release 10. 2. Redlands, CA: Environmental Systems Research Institute). A GIS vector
geospatial shapefile for Merlin nest sites was creating by importing and geo-referenced
coordinate data collected during field surveys. To calculate the distance to steams and open
water from each nest site the spatial join tool was used to combine vector stream and water body
data from the 2010 Natural Resources Canada National Hydro Network (NHN) and the 2014
USGS National Hydrography Dataset (NHD) with the nest site coordinate data. The distance
from nest sites to streams and open water was determined using the near function in the analysis
toolbox. Stream density was calculating using the line density function of the spatial analysis
toolbox. Topographic surface features were derived from a 2010 United States Geological
Survey 10 m resolution digital elevation model that covered the entire study area. The elevation
of the center point of analysis plots was extracted from a USGS using the extract values to points
function of the spatial analysis toolbox.
Land Cover Classification
5
National Land Cover Database 2006 (NLDC, 2006) land cover/ land use maps developed
by the Mutli Resolution Land Characteristics Consortium (MRLCC) and Natural Resources
Canada Earth Observation for Sustainable Development of Forests (EOSD) were used to create a
land cover map of the study area. The raster based maps were generated from medium resolution
(30 m²) geometrically rectified Landsat 7 Thematic Mapper (TM) satellite imagery collected in
2006. The maps were created by processing the satellite imagery using an unsupervised
classification and regression analysis (Franklin & Mulder, 2002; Lillesand, & Kiefer, 2000).
Additional data used to develop the groundcover analysis included high resolution color aerial
photographs, digital elevation models, and field data obtained from the U.S. Forest Service, and
Natural Resources Canada (O'Neil et al., 2006; Wulder & Nelson, 2002).
The derived data from NOAA and ESOD was combined and processed using ESRI
ArcMap 10.2 to create a 9 group habitat class map for the study area. The designated habitat
classes used for analysis were based on field assessments of land cover characteristics and
observed Merlin behavior, then correlated with the National Land Cover Database (NLCD)
classification system. (Drummond & Stillman, 2014; NLCD, 2006) See appendix A for
comprehensive descriptions of the habitat classes. In order to increase the accuracy of the
analysis, a mask was created to eliminate areas of open water, and areas above 1500 m in
elevation, the observed upper elevation limits of Merlin activity. The accuracy of the classified
image was verified using field gathered vegetation plot data as well as field data obtained from
United States Bureau of Land Management (BLM), and the British Columbia Ministry of Forests
(BC MOF). The classified image was resampled to a 25m² cell size in an effort to simplify
analysis rather than increase the accuracy. Prior to measuring landscape patterns, the spatial
analysis filter tool of ArcMap was used to perform low option 3x3 smoothing procedure of the
raster land cover map. The purpose of this step was to reduce the significance of anomalous cells
giving the map a greater relevance to natural landscape patterns.
Two types of metrics were analyzed to help define Merlin habitat: landscape composition
and landscape configuration (McGarigal & Marks, 1995). Composition refers to the abundance
of a land cover type or attribute, whereas configuration describes the spatial arrangement of
patches or features. Landscape composition characteristics of the habitat class map were
measured using the class metrics function of FRAGSTATS v4 spatial pattern analysis program
(McGarigal, Cushman, & Ene, 2012). Landscape pattern configuration variables for the study
were measured using the patch metrics function of FRAGSTATS.
Landscape scale habitat pattern characteristics were determined for areas within an 8km
radius (201.06 km²) plots centered on known nest sites. This area represents the observed extents
of Merlin's home range activity (Drummond & Stillman, 2014). 9 variables were utilized to
analyze habitat quality; 1) the total amount and percent of land cover contained in each of the 9
habitat classes, 2) patch density, a index of spatial heterogeneity; 3) the density of habitat edge
(m/ha), 4) the number of different habitat types within each 25m² cell defined habitat richness
(ha), 5) stream density (m/ha), 6) distance to riparian areas, 7) patch shape which indicates the
geometric complexity of the patch, 8) the amount of impervious surfaces, and 9) the percentage
of forest canopy cover.
6
Standardized residuals of each variable were tested for normality. Independent sample t-
tests were employed to test the distribution of habitat variables between “used” habitat plots and
“random” plots. Additional analysis of nest site density and habitat associations was conducted
using Pearson correlation coefficient testing. Statistical analysis testing was conducted using
SAS 9.3 (SAS Institute, Cary NC). Due to the conservation status of the Pacific Forest Merlin, an
alpha level of <0.05 was selected for all tests of significance (WDFW, 2013).
Ground Plots
Habitat microhabitat structure and configuration play a vital role in a land bird’s selection
of breeding territory and nest site (James & Shugart, 1970; Block & Brennan, 1993). In an effort
to depict the full spectrum of vegetative and physiographic characteristics reference 238
vegetation plots were collected during field assessments. The 25 m radius (0.196 ha) vegetation
plots were delineated and surveyed near Coastal Forest Merlin nest sites during the study period.
The classes defined in the process were designed to represent the various types habitat Coastal
Forest Merlin encounter in association with different aspects of their behavior.
RESULTS
Land Cover Characteristics
Analysis of habitat landscape configuration and composition was conducted on 70
individual non overlapping 8km radius (201.06 km²) home range plots centered on known
Merlin nest sites, and 33 randomly generated (Figure 3). In the 70 used home range plots the
largest habitat classes in terms of total area were conifer forest which comprised 3108.46 km²/
58.42 %; followed by; mixed forest 486.71 km²/ 9.15 %; shrub/scrub 412.90 km² / 7.76%,
agriculture 362.43 km²/ 6.35%, and developed low intensity 337.92 km²/ 6.14 %. The rest of the
habitat classes individually made up less than 5% of the groundcover in the analyzed regions
each (Table 1). Bare land constituted 0.38% of the land cover and was eliminated from analysis.
0
500
1000
1500
2000
2500
3000
3500
4000
4500
5000
Agriculture Deciduous
Forest
Developed Open
Space
Evergreen Forest High Intensity
Developed
Low Intensity
Developed
Medium Intensity
Developed
Mixed Forest Scrub/Shrub
Habitat Class
Sq
uare
Kil
om
ete
rs
Figure 3. Abundance and distribution of habitat classes within 70 201 km² analysis plots.
7
0
500
1000
1500
2000
2500
3000
3500
4000
4500
5000
Agriculture Deciduous
Forest
Developed Open
Space
Evergreen Forest High Intensity
Developed
Low Intensity
Developed
Medium Intensity
Developed
Mixed Forest Scrub/Shrub
Habitat Class
He
cta
res
Figure 4. Mean amounts and standard deviation of habitat classes within 20106 ha (201 km²) analysis plots.
Table 1. Description of classes used to characterize the breeding habitat of Coastal Forest Merlin.
Habitat Type Code Description
Developed
High Intensity HID
Industrial, commercial, and high density residential areas with 80-100% impervious surfaces.
These are zones cleared of major vegetation and include land cover such as concrete, tarmac, or
buildings.
Developed
Medium
Intensity
URES
Medium density residential, commercial and small city parks with 50-79% impervious surfaces.
Generally consist of areas of mixed deciduous/conifer forest, as well as non-native species.
Canopies have a low average percentage of vertical cover. Many non-native species are mixed
with natives trees to make a mixed, low density forest that usually have a minimum of 30%
conifer composition. Patches of vegetation are frequent but highly fragmented. The edge
interface between patches is generally complex in shape.
Developed
Low Intensity RRES
Rural areas with varying amounts of low density residential or commercial activity with 20-49%
impervious surfaces Contains groundcover that consists of a mix of shrub, herbaceous, and
forested areas of mixed species and seral development of conifers and hardwoods. These areas
have similar characteristics as developed medium density, but generally have smaller patch size
Developed
Open Space OPEN
Developed rural and urban areas that include; agriculture, pasture, grasslands, and utility
corridors. Generally have low height vegetation, and square edge configuration
Agriculture AGRI Areas of agricultural, pasture, and human created grassy area. Vegetation ≤ 2 m
Young Forest YOUNG
Early seral stage successional forests with an open or patchy canopy structure. Trees >5m tall,
>20% cover, >75% tree species shed foliage Usually riparian or disturbed areas with Red Alder,
Black Cottonwood, Big Leaf Maple, Vine Maple, conifer saplings and various shrubs.
Mixed Forest INTER
Areas of mixed forest in mid seral development with a patchy canopy closure; Trees >5m tall, >20% cover.
Contains a variety of species consisting of Douglas Fir, Western hemlock, Western red cedar, with some
Hardwoods as well. Arboretums, larger parks, some golf courses, and wooded reserves within areas of more
intense human activity.
Conifer Forest OGM
Mid and late stage seral conifer forest with a heterogeneous spatial configuration. Trees >5m
tall, >20% cover, >75% species maintain leaves. Typical species include Douglas Fir, Western
Hemlock, Western red cedar, grand fir, and Sitka spruce. These forests show a diversity of tree
ages and species, indicating a natural succession. Saplings to snags, young trees to trees
decaying from old age are represented. Patch size and core area tends to be relatively large.
Shrub/Scrub SCRUB
Non forested areas consisting of bare ground, small shrubs, saplings, or herbaceous groundcover
Shrubs/trees <5m tall, >20% cover. These areas consist of recent clear cuts, construction sites, or
areas of disturbance.
8
National Land Cover Database 2006 (NLCD200)
Coastal Forest Merlin Distribution and Density In Different Habitat Classes
The distribution and abundance of a Coastal Forest Merlin population was surveyed from
1986 to 2013. For the purposes of this project, a sample of 219 individual Coastal Forest Merlin
nest sites was used for analysis (Figure 3). Merlin nest sites were located in 4 of the 9 defined
habitat classes; conifer forest, mixed forest, developed medium intensity and developed low
intensity. The highest number of nest sites were located in the conifer forest class; 82, followed
by developed medium intensity 77, mixed forest; 37, and developed low intensity; 23.
The observed nest site densities were significantly higher in the developed medium
intensity and developed low intensity classes than Conifer and Mixed forest. Demographic
surveys revealed that the highest density of Merlin nest sites was located in the developed
medium intensity class which contained 77 of 219 nests or 35.16 % despite the class only
representing 577.08 km² or 4.10% of the available habitat in their home ranges. Similar
associations were noted in the developed low intensity class which contained 23 of 219 nest or
16.89% of 864.21km² or 6.14% of the available habitat.
Habitat Associations
Coastal Forest Merlin Preference
for habitat groundcover composition and
configuration as well as other geographic
and environmental factors was evaluated
by comparing the habitat characteristics
of used home range plots with randomly
generated home range plots. Refer to
appendix B for complete habitat variable
descriptions. In terms of topographic and
hydrographic features Coastal Forest
Merlins utilized locations that were
located in areas with a closer proximity
to areas with greater stream density
(TSD, t = 6.07, df = 105, P < 0.05),
closer to riparian areas (RIPA, Z = -5.59,
df = 105, P = 0.0003), and locations that
were lower in elevation (ELEV, t = 2.15,
df = 105, P < 0.05). In terms of landscape
configuration, Merlin sites were found
more frequently than in habitat classes
with a lower level of patch density
( PD, t = 3.22, df =105 , P = < 0.05),
greater edge density ( ED, t = 9.19, df =
105, P < 0.05), and greater complexity of
patch shape ( SHAPE, t = 2.12, df =105,
P < 0.05).
9
Nest site density had a moderately negative level of correlation for the amount of
impervious surfaces near nest sites (P = 0.22, r² = 0.41, ß = -0.64). Nest sites were located in areas
ranging from 0-75 % impervious ground cover. Nest site densities were not strongly correlated
with the percent forest cover (P = 0.67, r² = 0.14, ß = 0.35). 31 nest sites were located in areas with 5-
30% forest cover while areas with area with greater than 55% forest cover had consisted of 27
nest sites. Table 2. Comparisons of landscape characteristics for Coastal Forest Merlin study plots sampled in the study area. Values are means (± SD).
Landscape Habitat
Variable ª Plot Type t-test df =105
Used
( n =70 )
Random
(n = 37)
t P
Stream Dist. 290.04± 487.63 159.32±147.05 1.59 0.05
Riparian Dist. 498.65 ±845.13 954.19 ±1116.43 2.37 0.05
Stream Density 255.89 ± 306.01 238.42 ± 270.54 6.07 0.05
Elevation 194.86 ± 157.40 312.86 ±388.09 2.15 0.05
Edge Density 37.94 ±12.54 14.16 ±13.08 9.19 0.05
Patch Density 2.92±3.08 4.66±1.57 3.22 0.05
Patch Shape Complexity 1.77 ± .164 1.11 ± .298 2.12 0.05
Patch Shape SD 0.94 ± 0.20 0.72 ± 0 .26 4.87 0.05
Aggregation Index 71.23 ± 14.96 76.83 ± 10.26 2.12 0.05
Patch Size SD 60.35 ± 91.17 52.88 ± 102.53 1.78 0.05
Mean Patch Size 4.30 ± 6.04 1.28 ± .63 3.03 0.05
ª For descriptions see appendix A
Table 3. Outputs for Pearson Correlation coefficient calculations for Coastal Forest Merlin breeding habitat variables.
Landscape Habitat Variable
P r² ß
Level of
Correlation
Riparian Dist. 0.15 0.66 0.81
Negative Strong
Stream Density 0.04
0.75 0.86
Positive Strong
Edge Density 0.23
0.41 0.64
Positive Moderate
Patch Density 0.03
0.42 0.65
Positive Moderate
Patch Shape Complexity 0.17
0.53 0.73
Positive Moderate
Total Forest Cover 0.67
0.14 0.35
Positive Weak
Impervious Surfaces 0.22
0.41 0.64
Negative Moderate
10
CONCLUSIONS AND DISCUSSION
Land Cover Characteristics
The project used GIS to analyze important breeding habitat variables and associations for
a population of Coastal Forest Merlin in an 116,660 km² study area spread over Northwest
Washington State and British Columbia, Canada. The analysis did not attempting to predict
occurrence or distribution of Merlins in the study area. The goal was to depict similarities of
breeding habitat variables on a landscape level based on observed sightings of the species. By
using GIS to interface data layers that delineate land cover, land use, anthropogenic influence,
topography, and hydrography it was possible to characterize the study area in terms of the overall
value of each relevant habitat variable as well as their spatial distribution on the landscape. This
analysis method characterizes the habitat in the study area in terms of overall abundance of
habitat classes as well as the relative distributions on the landscape. Analysis determined that
quantifying land cover characteristics and configuration was an effective indicator of the
distribution and abundance of high quality Merlin habitat.
By reconciling the results of the analysis with field collected ground truth data, this
method demonstrated the ability to identify habitat characteristics for Pacific Forest Merlins at a
fine spatial scale over a large geographical area. This technique for habitat characterization can
be accomplished using readily available geospatial data combined with proper selection and use
of key species dependent habitat variables and an assessment of accuracy using reliable
validation data. The product of analysis can be used to delineate areas of high quality habitat for
a focal species, determine the amount of high quality habitat within a study area, and in and
evaluate different management plans designed to protect areas of valuable habitat.
The methods employed in this project could be used to perform analysis of landscape
associations for most species using a relevant set of key habitat variables. The methods could be
used to monitor changes in wildlife habitat before and after disturbances, or at different spatial
and temporal extents or resolution. Wildlife habitat analysis using GIS had the ability to expand
the awareness of habitat abundance, quality, and spatial configuration that would help guide the
resource management decision making process.
Habitat Associations
The spatial configuration of habitat variables and their association with a species
distribution and abundance is a becoming a commonly used method in landscape ecology
(Flather & Sauer, 1996). The analysis conducted for this project exemplified the association of
several key habitat variables on Coastal Forest Merlin and breeding success. Merlin frequently
use concealed perches located at forest edges to surprise and capture their prey, they may also
use their speed and agility to fly below the canopy in and flush prey. Some may use topographic
elements or landscape features to conceal their approach from potential prey (Johnsgard, 1990).
Merlins cache surplus food on a branch or in an unused nest located on a nearby tree (Cade,
1980). These aspects indicate two primary elements of quality foraging habitat are a fragmented
forest configuration with complex edge shape.
11
Breeding sites for Merlins were widespread across the study area and took place in a
variety of natural, semi natural, and development habitats. Merlins avoided using habitat near
areas with high levels of impervious surfaces and large expanses of open areas with abrupt
habitat edges. These relationships was negative with high intensity developed high intensity,
agricultural, and shrub/scrub habitat classes. These classes contained no nest sites and made up
little of the overall land cover in home ranges. Land cover analysis indicated highest Merlin nest
densities were linked lower levels of patch density indicating a preference for a fragmented
landscape common to areas of human development. Merlin home ranges consisted of a greater
variety of different habitat classes which demonstrates a preference for areas with a greater level
of habitat richness. Merlins density was linked with areas with complex patch shape with greater
amounts of habitat edge as indicated by a high level edge density. The association of higher
Merlin nest site densities with areas exhibiting these characteristics in greater amounts than was
what was generally available across the landscape indicates Merlins are demonstrating a
preference for these traits when selecting breeding territory.
Management Implications
The loss or alteration of habitat is an area of concern for the long term success of Coastal
Forest Merlin populations (Drummond & Stillman, 2014). Land cover in the study area
particularly in the study area had been increasing impacted by anthropogenic influences.
Although presently the natural and developed landscape of the study area provides Merlins with
an adequate amount of desirable breeding habitat, the increase of human population and the
resulting development has the potential to degrade or eliminate what currently exists. Current
methods of managing wildlife habitat are varied and diverse due to the different needs required
for multiple species (Rodiek & Bolen, 1991). Effective techniques for wildlife conservation are
seldom used; planning for wildlife habitat conservation is primarily conservative in approach,
and disjointed in application (McKinnon, 1987). As a result of the numerous plans exists that
were developed and managed by different agencies and have various levels of efficacy.
Numerous Federal, state, provincial, and local laws, ordinances, and special provisions exist that
may be used to conserve wildlife and their habitat. Many areas of habitat preferred by Merlins
for nest sites are slated for protection under a variety of regulatory acts (Rullman & Marzluff,
2014). Due to the fragmented approach to conservation, a regional approach to habitat
management may be the best method to help protect existing Merlin habitat. Many areas of high
quality habitat are located on private land not protected by existing management plans.
Incorporating outreach and education for private landowner into management plans may be an
effective method to protect these otherwise vulnerable areas.
ACKNOWLEGEMENTS
I express my sincere gratitude to David Drummond and Roger Stillman of the Merlin
Falcon Foundation who was enthusiastic, supportive, and informative as well as providing a high
level of support and useful feedback that was extremely beneficial to the quality of this project. I
would like to also thank my advisor Dr. Andrew J. Bach for his support and critiques during the
planning and revision stages of the process. And most importantly, I would like to thank my wife
and family for their feedback, patience and support.
12
Works Cited
Beer, J. R. (1966). The pigeon hawk in Minnesota. The Loon, 38(4) 129-132.
Block, W. M, and L. A. Brennan. (1993). The habitat concept in ornithology: Theory and applications. P.
35-91 In: D.M. Power (ed.). Current Ornithology, Volume
11. Plenum Press, New York.
Brown, L. and D. Amadon. (1968). Eagles, hawks, and falcons of the world. McGraw-Hill, New York.
Vol. 11:802-807.
Cade, T.J. (1982). Falcons of the world. Cornell University Press, Ithaca, NY.
COSEWIC. (2009). Canadian Wildlife Species at Risk. Committee on the Status of Endangered Wildlife
in Canada. Web site: http://www.cosewic.gc.ca/eng/sct0/rpt/rpt_csar_e.cfm [accessed 17 October
2011]
Dettmers, R., and Bart, J. (1999). A GIS modeling method applied to predicting forest songbird habitat.
Ecological Applications, 91:152-163.
Dickson, H.L., & Smith, A.R. (1991). Use of Landsat Thematic Mapper and multi-spectral scanning
imagery to identify habitats and shorebird nesting areas on the outer Mackenzie River Delta,
NWT. In: Marsh, P., and Ommanney, C.S.L., eds. Mackenzie Delta: Environmental interactions
and implications of development. NHRI Symposium No. 4.Saskatoon, Saskatchewan: National
Hydrology Research Institute, Environment Canada. 91 –106
Drummond, D. P.& Stillman, R.C. (2013). (In Review) Coastal Forest Merlin Breeding Habitat and
Climate Influence on Reproductive Success.
Earth Observation for Sustainable Development (EOSD). (2009). Land Cover, Circa 2000 – Vector.
Government of Canada, Natural Resources Canada, Earth Sciences Sector, Centre for
Topographic Information - Sherbrooke, Québec, Canada
Environmental Systems Research Institute (ESRI) (2013). ArcMap, version 10.2. Environmental Systems
Research Institute, Inc, Redlands, CA, USA.
Franklin, S.E., & Wulder, M.A. (2002). Remote sensing methods in large area land cover classification
using satellite data. Progress in Physical Geography, 26: 173–205.
Flather, C.H. & Sauer, J.R., (1996). Using landscape ecology to test hypothesis about large-scale
abundance patterns in migratory birds. Ecology, 77, 28-35.
Floberg, J., M. Goering, G. Wilhere, C. MacDonald, C. Chappell, C. Rumsey, Z. Ferdana, A. Holt, P.
Skidmore, T. Horsman, E. Alverson, C. Tanner, M. Bryer, P. Iachetti, A. Harcombe, B.
McDonald, T. Cook, M. Summers, D. Rolph. (2004).
Willamette Valley-Puget Trough-Georgia Basin Ecoregional Assessment, Volume One: Report.
Prepared by The Nature Conservancy with support from the Nature Conservancy of Canada,
Washington Department of Fish and Wildlife, Washington Department of Natural Resources
(Natural Heritage and Nearshore Habitat programs), Oregon State Natural Heritage Information
Center and the British Columbia Conservation Data Centre.
13
Fox, G. A. (1964). Notes on the western race of the pigeon hawk. The Blue Jay, 22(4): 140-147.
Franklin, S.E., & Wulder, M.A. (2002). Remote sensing methods in large area land cover classification
using satellite data. Progress in Physical Geography, 26: 173–205.
Fry, J., Xian, G., Jin, S., Dewitz, J., Homer, C., Yang, L., Barnes, C., Herold, N., and Wickham,
J.(2011). Completion of the 2006 National Land Cover Database for the Conterminous United
States, PE&RS, Vol. 77(9):858-864.
Glenn, E.M, & Ripple, W.J. (2004). On Using Digital Maps to Assess Wildlife Habitat. Wildlife Habitat
Mapping, 32(3):852-860, pp. 852-860.
Global Raptor Information Network. (2014). Species account: Merlin Falco columbarius. Downloaded
fromhttp://www.globalraptors.org on 21 Sep. 2014
Gratto-Trevor, C.L. (1995). Use of Landsat TM Imagery in Determining Important Shorebird Habitat in
the Outer Mackenzie Delta, Northwest Territories. Artic, VOL. 49, NO. 1 (March 1996) P. 11– 22
Homer, C.G., Edeards, T.C., Ramsey, R.D., & Price, K.P. (1993) Use of Remote Sensing In Modeling
Sage Grouse Winter Habitat. The Journal of Wildlife Management, Vol. 57, No. 1 (Jan., 1993),
pp. 78-84
Iachetti, P., J. Floberg, G. Wilhere, K. Ciruna, D.Markovic, J. Lewis, M. Heiner, G. Kittel, R.
Crawford,S. Farone, S. Ford, M. Goering, D. Nicolson, S. Tyler, and P. Skidmore. (2006). North
Cascades and Pacific Ranges Ecoregional Assessment, Volume 1 - Report. Prepared by the
Nature Conservancy of Canada, The Nature Conservancy of Washington, and the Washington
Department of Fish and Wildlife with support from the British Columbia Conservation Data
Centre, Washington Department of Natural Resources Natural Heritage Program, and
NatureServe. Nature Conservancy of Canada, Victoria, BC. James, F.C. & Shugart, H.H. Jr. (1970). A quantitative method of habitat description. Audubon Field
Notes 24, 727–736.
Johnsgard, P.A. (1990). Hawks, eagles, and falcons of North America. Smithsonian Institution Press,
Washington, D.C.
Johnston, R.M., & Barson, M.M. (1993). Remote sensing of Australian wetlands: An evaluation of
Landsat TM data for inventory and classification. Australian Journal of Marine and Freshwater
Research, 44:235–252.
Kittle, G., Cadrin, C., Markovic, D., Stevens, T. (2011). Central Interior Ecoregional Assessment:
Terrestrial Representation in Regional Conservation Planning. Journal of Ecosystems and
Management, North America, 12, may. 2011. Available at:
<http://jem.forrex.org/index.php/jem/article/view/103/58>. Date accessed: 22 Sep. 2014.
Knutson, K. L., and V. L. Naef. (1997). Management recommendations for Washington’s priority
habitats: riparian. Wash. Dept. Fish and Wildlife. Olympia. 181pp.
Lillesand, T.M. & Kiefer, R.W. (2000). Remote sensing and image interpretation, 4th edn. John Wiley
and Sons, New York.
14
McGarigal, K., Cushman, SA. and Ene, E. (2012). FRAGSTATS v4: Spatial Pattern Analysis Program
for Categorical and Continuous Maps. Computer software program produced by the authors at the
University of Massachusetts, Amherst
McGarigal, K. and. Marks, B.J. (1995). FRAGSTATS: spatial pattern analysis program for quantifying
landscape structure. Gen. Tech. Rep. PNW-GTR-351. Portland, OR: U.S. Department of
Agriculture, Forest Service, Pacific Northwest Research Station. 122 pp.
O'Neil, J. K., Kroll, C., Grob, C., Fassnacht, K., Alegria, J., Nighiberit, T., Demeo, T., Feirerma, J., &
Weiterman, D. (2000). Interagency vegetation mapping project (IVMP) Coastal Province final
release. United States Department of the Interior, Bureau of Land Management, and United
States Forest Service, Portland, Oregon, USA.
Rodiek, J.E. & Bolan, E.G. (1991). Wildlife and Habitats in Managed Landscapes. Island Press,
Washington D.C.
Rullman, S. & Marzluff, J.M. (2014). Raptor presence along an urban-woodland gradient: influences of
prey abundance and land cover. Journal of Raptor Research, 48(3): 257-272.
SAS Institute Inc. (2011). Base SAS® 9.3 Procedures Guide. Cary, NC: SAS Institute
Inc.
Shaw, D.M., & Atkinson, S.F. (1990). An introduction to the use of Geographic Information Systems for
ornithological research. Condor, 92:564-570.
Sodhi, N.S., Oliphant, L.W., James, P.C., & Warkentin, I.G. (1993). Merlin
(Falco columbarius). In The Birds of North America, No. 44 (A. Poole and F. Gill,
eds.).Philadelphia: The Academy of Natural Sciences; Washington DC: The American
Ornithologists’ Union.
Turner, M.G. & Gardner, R.H. (1991). Quantitative methods in landscape ecology: an introduction.
Quantitative methods in Landscape Ecology (ed. by M.G. Turner and R.H. Gardner), pp. 3–13.
Springer-Verlag, New York.
Tuttle, E.M., Jensen, R, J., Formica, V.V., and Gonser, R. J. (2006). Using remote sensing image texture
to study habitat use patterns: a case study using the polymorphic white-throated sparrow
(Zonotrichia albicollis). Global Ecology and Biogeography, 15, 349–357
Vander Schaaf, D., G. Wilhere, Z., Ferdaña, K., Popper, M., Schindel, P., Skidmore, D., Rolph, P.,
Iachetti, G., Kittel, R., Crawford, D., Pickering, L.and Christy, J. (2006). Pacific Northwest Coast
Ecoregion Assessment. Prepared by The Nature Conservancy, the Nature Conservancy of
Canada, and the Washington Department of Fish and Wildlife. The Nature Conservancy,
Portland, Oregon.
Washington Department of Fish and Wildlife. (WDFW) (2013). Threatened and Endangered Wildlife in
Washington: 2012 Annual Report. Listing and Recovery Section, Wildlife Program, Washington
Department of Fish and Wildlife, Olympia. 251 pp.
Weins, J.A. (1989). Spatial scaling in ecology. Functional Ecology, 3 385-397
15
Wulder, M.A., Nelson, T. (2003). EOSD Land Cover Classification Legend Report, version 2. Natural
Resources Canada, Canadian Forest Service, Pacific Forestry Centre, Victoria, British Columbia,
Canada, January 13, 2003. 83 http://www.pfc.forestry.ca/eosd/cover/EOSD_Legend_Report-
v2.pdf
APPENDIX A. Variable codes and explanations for landscape metrics used for analyses.
Variable Name Units Description Data Source
ED Edge Density m/ha Amount of edge relative to habitat
class area
NLDC, ESOD
2006 25m
SHAPE Patch Shape Complexity 0.0-
1.0
Greater complexity = ≥1
Less complexity= 1
NLDC, ESOD
2006 25m
PD Patch Density; Habitat
richness
#/ ha Number of different habitat patches
per ha
NLDC, ESOD
2006 25m
TCAI Total Core Area Index 0.0-
100.0
Percent of landscape containing core
area
NLDC, ESOD
2006 25m
SIZE Patch Size ha The mean area of patch size per
class
NLDC, ESOD
2006 25m
WATER Water 0.0-
100.0
Percent of open water in the
landscape
NHD/ NHN 2012
ELEV Elevation m Elevation of plot center USGS DEM 2010
25m
RIPA Distance to Riparian Zone m Distance to the boundary of nearest
riparian area
NHD/ NHN 2012
STRM Distance to Stream m Distance to nearest stream centerline NHD/ NHN 2012
TSD Total Stream Density m/ ha Total length of all streams in each
study plot in the study area
NHD/ NHN 2012
PIS Impervious Surface 0.0-
100.0
Percent of impervious surface in the
landscape
CCAP, ESOD
2006 25m
PFC Forest Cover 0.0-
100.0
Percent of forest cover in the
landscape
CCAP, ESOD
2006 25m
HID Developed High Intensity 0.0-
100.0
Percent of high intensity developed
area in the landscape
NLDC, ESOD
2006 25m
URES Developed Medium Intensity 0.0-
100.0
Percent of medium intensity
developed area in the landscape
NLDC, ESOD
2006 25m
RRES Developed Light Intensity 0.0-
100.0
Percent of light intensity developed
area in the landscape
NLDC, ESOD
2006 25m
MIXED Deciduous/Mixed Forest 0.0-
100.0
Percent -mixed forest in the
landscape
NLDC, ESOD
2006 25m
OGM Evergreen Forest 0.0-
100.0
Percent of conifer forest in the
landscape
NLDC, ESOD
2006 25m
SHRUB Shrub/ Scrub 0.0-
100.0
Percent of shrub/ scrub land in the
landscape
NLDC, ESOD
2006 25m
YOUNG Young Mixed Forest 0.0-
100.0
Percent of young mixed in the
landscape
CCAP, ESOD
2006 25m
AG Agriculture 0.0-
100.0
Percent of agriculture (pasture/hay)
in the landscape
CCAP, ESOD
2006 25m